使用ggplot与assign(paste0)在for循环中绘图时柱状图异常问题

问题现象 #

使用for循环基于AUX_GRAF数据框批量生成ggplot柱状图，生成的图表标题和geom_label显示正确，但所有图表的geom_bar都和最后一次迭代的柱状图一致。循环内打印图表显示正常，但循环外调用保存的图表时出现问题。

用户原代码：

for(i in colnames(AUX_GRAF[2: ncol(AUX_GRAF)])){
  AUX01 <- as.numeric(AUX_GRAF[, i])
  AUX <- ggplot (AUX_GRAF, 
                              aes(x = reorder(SE, as.numeric(SE)),
                                  y = AUX01)) + 
  theme(axis.text.x = element_text(face = "bold")) +
  labs(caption = "Fonte", 
       x = "Semana Epidemiológica",
       y = "Número de Casos",
       title = paste0(i, " - Notificados"))+
  theme( panel.grid.major = element_line(color = "#C0C0C0"),
         panel.grid.minor = element_blank(),
         panel.background = element_rect(fill = "#F5F5F5"),
         plot.title = element_text(face = "bold",
                                   size = 15,
                                   colour = "#556B2F")) +
  geom_bar(stat = "identity",
           color = "black",
           fill = "#8FBC8F") + 
  geom_label(aes(label = AUX01),
             alpha = 0.5,
             vjust = 0.1) +
  scale_y_continuous(expand = expansion(mult = c(0, 0.05)))
  assign(paste0("SE_HIST_NOT_", i), AUX)  
    Sys.sleep(1)
    print(assign(paste0("SE_HIST_NOT_", i), AUX))
    rm(AUX)
}

样本数据：

> dput(head(AUX_GRAF))
structure(list(SE = c("21", "22", "23", "24", "25", "26"), ARAPUÃ = c(2, 
3, 2, 1, 1, 0), ARIRANHA_DO_IVAÍ = c(0, 0, 0, 0, 0, 1), CÂNDIDO_DE_ABREU = c(1, 
0, 0, 1, 0, 0), CRUZMALTINA = c(0, 0, 1, 0, 0, 0), GODOY_MOREIRA = c(0, 
1, 0, 0, 0, 0), IVAIPORÃ = c(39, 32, 18, 16, 14, 10), JARDIM_ALEGRE = c(11, 
19, 7, 2, 0, 0), LIDIANÓPOLIS = c(22, 18, 9, 9, 4, 11), LUNARDELLI = c(12, 
2, 3, 3, 7, 2), MANOEL_RIBAS = c(5, 5, 1, 0, 3, 1), MATO_RICO = c(0, 
0, 0, 0, 0, 0), NOVA_TEBAS = c(15, 19, 11, 7, 3, 5), RIO_BRANCO_DO_IVAÍ = c(10, 
1, 0, 0, 0, 0), ROSÁRIO_DO_IVAÍ = c(0, 0, 0, 0, 0, 0), SANTA_MARIA_DO_OESTE = c(0, 
0, 0, 0, 0, 0), SÃO_JOÃO_DO_IVAÍ = c(3, 2, 0, 0, 0, 1)), row.names = c(NA, 
6L), class = "data.frame")

问题原因 #

ggplot的aes()参数采用延迟求值机制：ggplot对象不会立即计算aes中变量的具体值，而是在实际渲染（如print()或保存图表）时才去读取变量的当前值。

在原代码中，y = AUX01引用的是循环外部的变量，循环结束后AUX01保留的是最后一次迭代的数值，因此所有保存的ggplot对象在渲染时都会读取这个最终值，导致柱状图全部显示为最后一次迭代的结果。

虽然geom_label(aes(label = AUX01))看似正常，但实际上也存在同样的延迟求值风险，只是当前场景下未暴露。

解决方案 #

有两种可靠的解决方法：

方法1：创建局部数据框（推荐） #

每次迭代时生成包含当前列数据的局部数据框，确保ggplot绑定的是当前迭代的数据集：

library(dplyr)

for(i in colnames(AUX_GRAF[2: ncol(AUX_GRAF)])){
  # 生成当前迭代的局部数据，包含排序后的SE和当前列数值
  local_data <- AUX_GRAF %>%
    mutate(
      SE_ordered = reorder(SE, as.numeric(SE)),
      current_cases = as.numeric(.[[i]])
    )
  
  # 基于局部数据绘图
  AUX <- ggplot(local_data, aes(x = SE_ordered, y = current_cases)) +
    theme(axis.text.x = element_text(face = "bold")) +
    labs(
      caption = "Fonte",
      x = "Semana Epidemiológica",
      y = "Número de Casos",
      title = paste0(i, " - Notificados")
    ) +
    theme(
      panel.grid.major = element_line(color = "#C0C0C0"),
      panel.grid.minor = element_blank(),
      panel.background = element_rect(fill = "#F5F5F5"),
      plot.title = element_text(face = "bold", size = 15, colour = "#556B2F")
    ) +
    geom_bar(stat = "identity", color = "black", fill = "#8FBC8F") +
    geom_label(aes(label = current_cases), alpha = 0.5, vjust = 0.1) +
    scale_y_continuous(expand = expansion(mult = c(0, 0.05)))
  
  # 保存并打印图表
  plot_name <- paste0("SE_HIST_NOT_", i)
  assign(plot_name, AUX)
  print(AUX)
  rm(AUX)
}

方法2：使用.data代词直接引用列名 #

利用ggplot的.data代词直接引用当前列名，强制ggplot绑定到数据框的对应列，避免延迟求值问题：

for(i in colnames(AUX_GRAF[2: ncol(AUX_GRAF)])){
  AUX <- ggplot(AUX_GRAF, aes(
    x = reorder(SE, as.numeric(SE)),
    y = .data[[i]]  # 用.data代词直接引用当前列
  )) +
    theme(axis.text.x = element_text(face = "bold")) +
    labs(
      caption = "Fonte",
      x = "Semana Epidemiológica",
      y = "Número de Casos",
      title = paste0(i, " - Notificados")
    ) +
    theme(
      panel.grid.major = element_line(color = "#C0C0C0"),
      panel.grid.minor = element_blank(),
      panel.background = element_rect(fill = "#F5F5F5"),
      plot.title = element_text(face = "bold", size = 15, colour = "#556B2F")
    ) +
    geom_bar(stat = "identity", color = "black", fill = "#8FBC8F") +
    geom_label(aes(label = .data[[i]]), alpha = 0.5, vjust = 0.1) +
    scale_y_continuous(expand = expansion(mult = c(0, 0.05)))
  
  # 保存并打印图表
  plot_name <- paste0("SE_HIST_NOT_", i)
  assign(plot_name, AUX)
  print(AUX)
  rm(AUX)
}

说明 #

两种方法都能解决延迟求值导致的问题，其中方法1通过局部数据框明确隔离每次迭代的数据，逻辑更清晰；方法2更简洁，适合不需要额外数据处理的场景。

内容的提问来源于stack exchange，提问作者Gustavo